In the World of AI: Week of 2 June 2026, Australia Joins the Mythos Cyber Defence Program

Cyber defence, memory and the phone in every staff member's pocket all changed this week. Anthropic widened access to its most capable security model to include Australian organisations, OpenAI changed how ChatGPT remembers, and Apple showed a rebuilt Siri that can hand work to rival assistants. Here is the week to 9 June 2026.

1. Australia is included in Anthropic's Mythos Preview expansion

iTnews reported on 3 June that Anthropic is expanding Project Glasswing, its early access program for Claude Mythos Preview, to up to 150 new organisations in more than 15 countries, and that Australian organisations are included. A spokesperson would not name them. Access is tightly scoped to organisations defending critical infrastructure or building and operating systems in sectors such as power, water, healthcare, communications and financial services. Anthropic also said it expects other AI companies to release Mythos-class models and is concerned about such models being made public without safeguards against misuse.

A Conversation article republished on 4 June argues the benefits will flow mainly to large organisations and that smaller companies may lack the capacity to use such tools or to respond to a rising volume of security updates. It says the Australian Signals Directorate welcomed the access. Its analysis suggests businesses prepare for faster vulnerability discovery, build the skill to tell real vulnerabilities from false reports, and take care with AI systems tied to critical functions because they can be manipulated.

Why it matters: Better defence at big firms could mean fewer breaches that expose your customers' data. The same article says a growing stream of fixes will fall on smaller businesses.

This week: Turn on automatic updates wherever you can, and name one person who checks that critical systems have received them.

2. ChatGPT gets a new memory system

On 4 June 2026 OpenAI introduced a rebuilt memory system for ChatGPT, according to iClarified. It learns from past conversations, aims to drop out-of-date details and adds a summary page where users can review, add or correct what ChatGPT knows. Plus and Pro users in the United States received it first, free users follow over the coming weeks, and OpenAI plans to expand to more countries and to ChatGPT Go users.

Why it matters: If a staff member uses a personal ChatGPT account for work, details about clients or projects may now persist in that account's memory. The report does not describe retention settings or how to delete memories, so find out for yourself before relying on it. Our answer on stopping staff pasting confidential data into AI covers the rule to set.

This week: Ask staff who use AI tools for work whether they use a personal or business account, and move client work to a business account.

3. Apple unveils a rebuilt Siri and swappable assistants

At WWDC, Apple unveiled a rebuilt Siri on 8 June 2026, according to The Next Web. Siri becomes a standalone chatbot app that can search messages, emails and photos and carry out multi-step tasks across apps. Requests split across three tiers, from on-device processing to Apple's Private Cloud Compute to Google Cloud for the heaviest reasoning. Apple says queries are processed statelessly and not retained, and the article says Apple's contract bars Google from training future models on Apple user data. Federighi is quoted as saying "Data is only used to execute your request." iOS 27 Extensions let users set a third-party assistant such as Claude or ChatGPT as their default. Developer betas are out, public betas are expected in mid-July and general release is planned for autumn. The article does not say which countries will get Siri AI.

Why it matters: Staff may soon pick an assistant on their own phone that can see their email and messages, including work accounts. That is what shadow AI looks like in practice.

This week: Add one line to your staff guidance about which assistants may touch work email and messages, and review it when iOS 27 ships. Our guide to writing an AI use policy shows how.

What to do this week

• Turn on automatic updates and name the person who checks them.
• Find out which staff use personal AI accounts for work and move client work to a business account.
• Add a line on phone assistants and work email to your staff guidance.

Sources

• iTnews, "Anthropic opens Claude Mythos Preview AI program to Australia", 3 June 2026: itnews.com.au
• The Conversation (republished by the University of Melbourne Find an Expert), "Australia now has access to Anthropic's Claude Mythos. It may improve cyber safety, but not for everyone", 4 June 2026: findanexpert.unimelb.edu.au
• iClarified, "OpenAI launches Dreaming V3 memory system for ChatGPT", 4 June 2026: iclarified.com
• The Next Web, "Apple finally ships its AI do-over: Siri AI, a standalone app, and a three-tier privacy stack", 8 June 2026: thenextweb.com